Our understanding of schizophrenia (SCZ), bipolar disorder (BPD) and anorexia nervosa (AN) is advancing rapidly. We have identified polymorphisms and genes associated with all three disorders, although AN is still understudied compared to SCZ and BPD. As sample sizes for genome-wide association studies increase, larger numbers of associated variants will surely be identified, particularly for AN, which is projected to increase to 50,000 cases from ~3,500 currently, by 2019. However, such studies provide, at best, long lists of associated loci, which are not easily biologically interpretable. Consequently, we do not yet understand the key biological mechanisms underlying these diseases, and few effective treatments or medications are available. Methods that provide insight into the associations from these studies will be vital to furthering our understanding of disease etiology, and will have substantial public health impacts. We propose to develop statistical models to translate existing associations from these studies into biologically relevant information. These models are an innovative approach that capitalize on existing successful genetic studies. We use large, publicly available ?multi-omic? datasets with proven relevance to SCZ, BPD, and AN (for example brain gene expression, cell-type specific histone modifications, and gut microbiota) to build powerful multi-omic predictors. These may be used to predict higher-level measures (for example gene expression) from genotype, and test for association with disease. These types of associations may lead to increased understanding of underlying biological mechanisms, and opportunities for development of medications and therapeutic interventions. In specific aim 1, we will update and improve on our existing brain gene expression prediction models, using a large collection of post-mortem brain samples from the dorso-lateral pre-frontal cortex and anterior cingulate cortex. These samples will allow us to build large, well-powered, highly accurate prediction models. We will apply these models to existing studies of SCZ, BPD, and AN to provide disease-associated genes. In specific aim 2, we will extend our approach to include prediction of developmental brain gene expression, and again will apply our models to studies of SCZ, BPD, and AN. These analyses will provide trajectories of gene expression throughout development, and will identify genes associated with SCZ, BPD and AN at distinct developmental stages. In specific aim 3, we will create models predicting cell-type specific histone modifications and gut microbial composition from genotype, and will apply these to studies of SCZ, BPD, and AN. These analyses will elucidate the role of specific histone modifications (H3K4me3 and H3K27ac), in neurons and non-neurons, as well as the role of microbial diversity and specific bacterial species, in SCZ, BPD, and AN.